Epigenetic mechanisms underlying cognitive impairment and Alzheimer disease hallmarks in 5XFAD mice

5XFAD is an early-onset mouse transgenic model of Alzheimer disease (AD). Up to now there are no studies that focus on the epigenetic changes produced as a result of Aβ-42 accumulation and the possible involvement in the different expression of related AD-genes. Under several behavioral and cognition test, we found impairment in memory and psychoemotional changes in female 5XFAD mice in reference to wild type that worsens with age. Cognitive changes correlated with alterations on protein level analysis and gene expression of markers related with tau aberrant phosphorylation, amyloidogenic pathway (APP, BACE1), Oxidative Stress (iNOS, Aldh2) and inflammation (astrogliosis, TNF-α and IL-6); no changes were found in non-amyloidogenic pathway indicators such as ADAM10. Epigenetics changes as higher CpG methylation and transcriptional changes in DNA methyltransferases (DNMTs) family were found. Dnmt1 increases in younger 5XFAD and Dnmt3a and b high levels in the oldest transgenic mice. Similar pattern was found with histone methyltransferases such as Jarid1a and G9a. Histone deacetylase 2 (Hdac2) or Sirt6., both related with cognition and memory, presented a similar pattern. Taken together, these hallmarks presented by the 5XFAD model prompted its use in assessing different potential therapeutic interventions based on epigenetic targets after earlier amyloid deposition.

INTRODUCTION two decades to elucidate the remaining genetic risk for AD [12]. However, none of these candidates has proven to consistently influence disease risk or onset age in more than a handful of samples [13]. Reasons for lack of reproducibility may include the following: insufficient study power to detect variants with minor contributions; biologic, genetic, and allelic heterogeneity; difference in study design, and the presence of population substructure [14]. The occurrence of phenotypical differences in monozygotic twins over time is thought to arise from epigenetic changes induced by different environments or by stochastic events [15]; thus, it appears plausible that epigenetic changes resulting in altered gene expression may also be involved in the pathogenesis of LOAD.
In addition to genetic factors, growing evidences demonstrated that the effects of environmental factors and epigenetic mechanisms also play an important role in the pathogenesis of AD [16][17][18][19][20]. Epigenetic information is principally encoded in two types of synergistically acting covalent modifications: DNA methylation and chromatin modifications; these changes afford the epigenetic complexity that directly affects regulation of the gene expression and other genomic functions [21]. For example, hypomethylation of the Amyloid Precursor Protein (APP) gene in the brain of a patient with AD was found [22]. Subsequently, several studies have linked DNA methylation changes in hippocampus and cortex of patients with AD, showing that DNA hypomethylation is associated with a huge number of Amyloid β (Aβ) plaques [22]. DNA methylation is also involved in APP processing and Aβ production through the regulation of Presenilin1 (PS1) and β-secretase (BACE) expression [23]. Monitoring the frontal cortex of patients with AD, indicated that DNA hypomethylation leads to expanded regulation of proinflammatory genes, Nuclear Factor-κβ (NF-κβ) and the encoding of cyclooxygenase-2 that catalyzes the production of prostaglandins and some prostanoids [24,25]. This assumes that abnormal DNA methylation may cause nerve inflammation. Hypermethylation of Brain Derived Neurotrophic Factor (BDNF) and cAMP Response Element Binding (CREB) promoters would harm the plasticity of the synapses [25]. Although these studies report AD brain global DNA methylation changes in opposite directions, differences are likely due to the fact that the authors studied different brain regions and employed different methylation detection techniques.
In addition, Aβ changes in DNA methylation are associated with induction of several genes, causing apoptotic and cell loss. It remains unclear how these changes in DNA methylation began, but it is suggested that various cellular mechanisms, such as OS and Aβ-42, are involved [26]. OS and Aβ-42 may modify DNA methylation, and evidences for the dynamic modulation of learning and memory suggest the functional importance of abnormal methylation of DNA in the cases of AD [27].
The detection and identification of mutations in APP, PS1 and PS2 genes associated with Familial AD (FAD) have allowed the generation of various transgenic murine models. These transgenic models have the potential to enable us to understand the pathogenesis of the disease, and to develop and evaluate new and effective therapeutic targets. The 5XFAD mouse was formally established and declared in 2006 by Oakley and the specific characteristic of this model being that it is generated to promote rapid, aggressive and complete development of AD pathological interactions. This novel model of Aβ pathology consists of two APP mutations (APP K67ON/ M671L (Swedish), I716V (Florida), and V717 (London) and two PS1 mutations (PS1 M146 and L286V) on B6/SJL and each transgene is regulated by Thy1 promoter [28,29].
The present work focuses on the study of the possible involvement of Aβ-42 accumulation in the production of OS and in subsequent modifications in the epigenetic machinery in the murine model of AD, 5XFAD. We aimed to discuss the role of global DNA methylation and the histone epigenetic enzymes involved in epigenetic modification acting in the control of gene expression and memory in this AD mouse model.

5XFAD showed cognitive impairment but no changes in emotional parameters with age
NORT analysis demonstrated significant differences between the 5XFAD and Wt strains at 8, but not at 2 months of age (age, F(1,28) = 4.5631, p = 0.0416; genotype, F(1,28) = 4.563, p = 0.001, and interaction age × genotype F(1,28) = 6,211, p = 0.0189). The DI was significantly decreased in 5XFAD mice at 8 months of age compared with young 5XFAD mice and with age-matched Wt, indicating impairment in hippocampal memory processes (Fig. 1A). Although, all mouse groups were able to learn through trial days in the MWM paradigm, statistical differences were found comparing young 5XFAD compared with old Wt. (Fig.  1B). On the test day, it was observed that 5XFAD mice aged 2 and 8 months showed impairment in spatial memory, with a significant decrease in time spent in the platform quadrant and with erratic preference for the platform area (age, F (1,28) = 0.8253, p = 0.3720; www.impactaging.com genotype, F(1,28) = 19.67, p = 0.0001, and interaction age × genotype F(1,28) = 0.3977, p = 0.5334) (Figs. 1C and 1D).
Quantification of fluorescence intensity demonstrated higher astrogliosis in hippocampus of 5XFAD with regard to Wt and significant worsening with age (See Figs. 6B-D for statistical notations). Moreover, signifi-       (Fig. 8F).

DISCUSSION
Epigenetic modifications such as global DNA methylation/hydroxymethylation, histone demethylation levels and histone acetylation levels play a crucial role in gene regulation [30]. Epigenetic changes could be the basis for the presentation of senescence and for neurodegenerative diseases, such as AD [31]. In fact, impaired global DNA methylation has been found in aged and AD brain, but in addition other changes can be related with AD. For example, AD brain showed hyperand hypomethylated CpG islands in promoter regions for CREB and NF-kB genes, respectively. Moreover, AD brain demonstrated increased global histone H3 acetylation and hypermethylation of the promotor region for the drebrin-like protein gene and many other epigenetic changes observed were inversely related with respective changes in messenger RNA and protein levels [25]. Because OS may be induced by Aβ, it can be an important contributor to DNA epigenetic modifications. It has been shown that ROS can command epigenetic modification by histone acetylation and deacetylation by HATs and HDACs, respectively, linking OS to chromatin remodeling [32]. However, the exact relationship among Aβ, ROS and epigenetics remains largely unknown, in terms of whether Aβ does, in fact, cause DNA methylation/demethylation or whether it is responsible for other epigenetic changes that exert an influence on the progression of AD. 5XFAD mice are characterized by higher levels in amyloidogenic pathology [33]. These mice accumulate high levels of intraneuronal Aβ-42 at around 1.5 months of age with amyloid deposition rapidly following at around two months of age, first in the subiculum and layer 5 of the cortex and increasing rapidly with age [34]. Plaques spread throughout the hippocampus and cortex by 6 months of age. Gliosis also begins at around 2 months of age, developing in parallel with plaque deposition. Synapse degeneration is also observed (at approximately 4 months of age) as well as neuronal loss and deficits in spatial learning (at approximately 4 to 5 months of age) [35]. Tangles are not typical in this model but tau hyperphosphorylation has been described www.impactaging.com [36,37]. We demonstrated, as expected, that in our hands female 5XFAD exhibited an increase in AD markers with age, including, an increase in APP and Bace1 expression, higher protein levels of BACE1 and sAPPβ, and higher number of amyloid plaques indicating the development of AD pathology as described elsewhere in aged 5XFAD [38]. Another hallmark demonstrated in these transgenic animals comprised the increase in oxidative environment that parallels with the degree of disease development [39]. In our hands, oxidative stress initiated with 4-HNE increases at 2 months of age, but a complete development of oxidative markers was found at older ages, accordingly to the higher levels of amyloid pathology and also of tau hyperphosphorylation. AD brain undergoes a general OS process, as documented by significant protein oxidation, lipid peroxidation, and DNA oxidation [40]. Interestingly, 4-HNE, a reactive end-product of lipid peroxidation that is found increased in AD brains, has been observed to bind to histone lysine residues.
These molecular and biochemical findings are phenotypically demonstrated by loss in memory abilities and also by impairment in behavioral and emotional parameters in aged 5XFAD. Therefore, in our hands female 5XFAD initiate the AD characteristics as earlier as 2 months but the disease is completely developed at 8 months of age in a similar way described by Oakley and coworkers (2006).
Here we confront the implication of epigenetic changes with AD hallmarks in female 5XFAD (including Aβ deposition and OS processes), which we addressed from a holistic point of view, in order to demonstrate a possible relationship between the processes gated to earlier Aβ deposition that account for, in the 5XFAD brain, the epigenetic response, both to delay or to contribute the evolution of the disease.
By means of this approach, we demonstrate that a significant increase in global DNA methylation, measured by 5-methylCytosine (5-mC) levels in young animals when Aβ initiate deposition and with concomitant increase of Dnmt1 and Dnmt3a. Dnmt3b is maintained as increased in older mice. Although no changes in hydroxymethylCytosine (5-hmC) levels were found, Tet1 but not Tet2 (which catalyze the conversion of 5-mC into 5-hmC) increases in 2 month-old transgenic mice, followed by a reduction in older animals. In reference to histone modification by acetylation or deacetylation, again we found differences in transgenic mice, with increases in Hdac2, Jarid1a or G9a.
It was described elsewhere that DNMT3a and DNMT3b are essential for memory formation and that they participated in neuronal and synaptic plasticity changes. Recently, it is has been reported that DNMT3b moderates cognitive decline in subjects with mild cognitive impairment [41,42]. The significant changes found in Dnmt1 and Dnmt3b in young 5XFAD lead to The values used to calculate Partial correlation controlling for group coefficients were Amyloid pathology obtained with S-Thioflavin staining (showed in Figure 7B); behavioral parameters from NORT, EPM and OFT (showed in Figures 1A, 2A and 3C); protein expression for SOD1 and 4-HNE (showed in Figure 5A and 5B); and gene expression for Dnmt1 and Hdac2 (showed in Figures 8C and  9D) Correlation(2-tailed) is significant * p<0.05; **p<0.01 and ***p<0.001; (-) Negative covariation of two variables.
www.impactaging.com higher cytosine methylation, prior to the development of AD hallmark in this mouse, underlying the earlier participation for methylation processed at the onset of the pathology. Some authors reported that in aged mice there occurs a diminution in hippocampal expression of Dnmt3a, and that increasing this expression reverses memory deficits whereas knockdown in young mice impairs memory formation [43]. The increases in Dnmt3a and Dnmt3b in transgenic animals could be an attempt to surpass the harmful effect of the Aβ increment as well as the high oxidant environment that occurs in 5XFAD when the disease is completely developed, as depicted in negative correlations found among the results obtained in the present work (Table 4).
JumonjiC (JmjC) and ARID domain, containing histone lysine demethylase 1a (JARID1a) increased histone acetylation by inhibiting histone deacetylase 1 function and modifying gene transcription in a demethylaseindependent manner. Jarid1a and Jarid1b catalyzed H3K4 demethylation contributing to the silencing of retinoblastoma target genes in senescent cells. Therefore Jarid1a and Jarid1b are tumor-suppressors that control cellular senescence [44]. Moreover, depletion of JARID1a in mammalian cells reduced per histone acetylation promoter, dampened gene expression and, for example, shortened the period of circadian rhythms [45].
Different histone methylations are also associated with behavioral disorders. G9a methylates H3K9 and this hypermethylation leads to the downregulation of BDNF [46]. It well established that BDNF plays an important role in neuron formation, maintenance and memory formation, and memory formation affects acetylation of H3 and H4 N-terminal tails. The lowest level of G9a gene expression in 5XFAD compared with nontransgenic mice is coincident with the loss of memory in this AD mouse model.
Finally, histone modification also plays an important role in epigenetic control. Histone deacetylases have been implicated in memory and cognition, and concretely, HDAC2 plays a central role in coupling lysine acetylation with synaptic plasticity and its modulation is implicated in cognition and disease [47].
Sirtuins are a family of nicotinamide adenine dinucleotide-dependent deacetylases that are implicated in a number of physiological and pathological processes, such as control of glucose and lipid metabolism, cancer, genomic stability and DNA repair [48]. SIRT6 plays a role in synaptic function and neuronal maturation and may be implicated in the regulation of neuronal survival [49] and the overexpression of the SIRT6 extended life span [50]. It is noteworthy that, in the 5XFAD model, Hdac2 increased significantly in 2-month old transgenic mice whereas diminution of Sirt6 was found. No significant changes were found for these and other histone acetylases or deactetylases tested at older age (Hdac1, Sirt1 and Sirt2). A very significant negative correlation between amyloid plaques and HDAC2 was found (Table 4), supporting the previously reported role of HDAC2 in synaptic plasticity [51] and working memory results obtained in 5XFAD.
Our results support that the Aβ pathology that is characteristic for 5XFAD mouse correlates with behavior, OS markers and specific epigenetic enzymes in 5XFAD (Table 4). Depending on disease stage development, DNA transcription modification processes, such as methylation or acetylation, may form part of a vicious cycle involving OS because the increase in Aβ accumulation, and that of, in turn, Aβ may induce global DNA changes in methylation and acetylation that correlated with the progression of the pathology from earlier onset to final stages of the disease in 5XFAD, including neuronal loss, gliosis and the disturbances in cognition and behavior present in this animal model of AD.
Studies were performed in accordance with the Institutional Guidelines for the Care and Use of Laboratory Animals established by the Ethical Committee for Animal Experimentation at the University of Barcelona.

Behavioral and cognitive experiments
Elevated Plus Maze test. This test is employed to assess anxiety [52]. The basic measurement is the animal's preference for dark, enclosed over bright, exposed places. The animal is placed in the center of the apparatus and observed for a set time. Open Field Test. Open Field Test (OFT) apparatus was constructed of white plywood (50 × 50 × 25 cm) [53]. Red lines were drawn on dividing the floor into 25-cm squares. Behavior was scored with SMART ® ver. 3.0 software, and each trial were recorded for later analysis, utilizing a camera fixed to the ceiling at a height of 2.1 m situated above the apparatus. Mice were placed at the center, or at one of the four corners, of the Open Field and allowed to explore the apparatus for 5 min. After the 5-min test, the mice were returned to their home cages and the Open Field was cleaned with 70% ethyl alcohol and allowed to dry between tests. To assess the animal's habituation process to the novelty of the arena, the mice were exposed to the apparatus for 5 min on 2 consecutive days. The behaviors scored included Line Crossing, Center Entries, Center Stay Duration, Rearing, Defecation, and Urination. Each animal was then given a score for total locomotor activity, which was calculated as the sum of total distance, line crosses and number of rears.
Novel Object Recognition Test (NORT). This test was conducted in a 90-degree, two-arm, 25-cm-long, 20-cmhigh maze [54]. Light intensity in the middle of the field was 30 lux. The objects to be discriminated were plastic figures (object A, 5.25-cm high, and object B, 4.75-cm high). First, mice were individually habituated to the apparatus for 10 min for 3 days. On day 4, they were submitted to a 10-min acquisition trial (first trial), during which they were placed in the maze in the presence of two identical novel objects (A+A or B+B) placed at the end of each arm. A 10-min retention trial (second trial) occurred 2 h later. During this second trial, objects A and B were placed in the maze, and the times that the animal took to explore the new object (tn) and the old object (to) were recorded. A Discrimination Index (DI) was defined as (tn-to)/(tn+to). In order to avoid object preference biases, objects A and B were counterbalanced so that one half of the animals in each experimental group were first exposed to object A and then to object B, whereas the other one half first saw object B and then object A was presented. The maze, the surface, and the objects were cleaned with 96º ethanol between the animals' trials to eliminate olfactory cues.
Morris Water Maze test. An open circular pool (100 cm in diameter, 50 cm in height) was filled halfway with water [55] and water temperature was maintained at 22°C ± 1. Two principal perpendicular axes were defined; thus, the water surface was divided into four quadrants (NE, SE, SW, and NW), and five starting points were set (NE, E, SE, S, and SW). Four visual clues were placed on the walls of the tank (N, E, S, and W). Non-toxic white latex paint was added to make the water opaque, and a white escape platform was submerged 1 cm below the water level (approximately in the middle of one of the quadrants).
The animal's swimming paths were recorded by a video camera mounted above the center of the pool, and the data were analyzed with SMART ® ver. 3.0 statistical software. The learning phase consisted of 6 days of trials for each mouse. The animals were submitted to five trials each day starting from the positions set (in random order) and without a resting phase between each trial and the subsequent one. At each trial, the mouse was placed gently into the water, facing the wall of the pool, and allowed to swim for 60 sec. If not able to locate the platform in this period of time, the mouse was guided to the platform by the investigator. Animals were left on the platform each time for 30 sec in order to allow spatial orientation.
The parameters measured were latency time in finding the platform, time spent in each quadrant, and distance swum for each trial; the mean was calculated for each trial day. A memory test was performed at the end of the learning days, in which the platform was removed and the time spent by each mouse in each quadrant was measured.
Brain isolation and immunoanalysis assays. Mice were euthanized 3 days after the last trial was conducted, and the brain was quickly removed from the skull. Hippocampus were dissected and frozen in powdered dry ice and maintained at −80°C for further use. Tissue samples were homogenized in lysis buffer containing phosphatase and protease inhibitors (Protease Inhibitor Cocktail 2, Sigma), and cytosol and nuclear fractions was obtained as described elsewhere. Protein concentration was determined by the Bradford method. Fifteen μg of protein was separated by Sodium Dodecyl Sulfate-PolyAcrylamide Gel Electrophoresis (SDS-PAGE) (8-15%) and transferred onto PolyVinylidene DiFlouride (PVDF) membranes (Millipore). The membranes were blocked in 5% non-fat milk in Tris-Buffered Saline containing 0.1% Tween 20 (TBS-T) for 1 h at room temperature, followed by overnight incubation at 4°C with primary antibodies diluted in TBS-T and 5% Bovine Serum Albumin (BSA) (See details in Table 1). Membranes were then washed and incubated with secondary antibodies for 1 h at room temperature. Immunoreactive proteins were visualized www.impactaging.com utilizing an Enhanced ChemiLuminescence-based detection kit (ECL Kit; Millipore) and digital images were acquired employing a ChemiDoc XRS+System (BioRad). Band intensities were quantified by densitometric analysis utilizing Image Lab software (BioRad) and values were normalized to GAPDH.
For immunohistochemical studies, the frozen brains were embedded into OCT Cryostat Embedding Compound (Tissue-Tek, Torrance, CA, USA), cut into 20-μm-thick sections on a cryostat (Leyca Microsystems, Germany) at −20°C, and placed on slides. After 3 h of drying time at room temperature, the slices were fixed with acetone at 4°C for 10 min, allowed to dry overnight, and finally stored at -20°C until their further staining. For the staining procedure, the brain sections were first rehydrated by 5-min incubation in Phosphate-Buffered Solution (PBS). Afterward, the blocking/permeabilization step was performed (20 min in PBS 1% BSA + 1% Triton). Following two, 5-min washings in PBS, the slices were incubated overnight at 4°C with the primary antibodies (see Table 1 for dilutions). Two further washings were carried out prior to incubation with the fluorescent secondary antibody (1 h at room temperature, see Table for dilutions). Finally, before mounting with Fluoromount-G ™ (EMS, Hatfield, NJ, USA), nuclear staining was performed with Hoechst 2 µg/mL for 5 min at room temperature. Slides were allowed to dry overnight after mounting and image acquisition was performed with a fluorescence laser microscope (Olympus BX41; Germany). At least four images from four different individuals by group were analyzed with ImageJ/Fiji software available online from the National Institutes of Health, NIH). Thioflavin S staining. Slides were allowed to defreeze at room temperature and then were rehydrated with PBS solution for 5 min. Later, the brain sections were incubated with 0.3% thioflavin S (Sigma-Aldrich) for 20 min at room temperature in the dark. Subsequently, these were submitted to washes in 3-min series, specifically with 80% ethanol (two washes), 90% ethanol (one wash), and three washes with PBS. Finally, the slides were mounted using Fluoromount (EMS), allowed to dry overnight at room temperature in the dark, and stored at 4ºC. Image acquisition was performed with an epifluorescence microscope (BX41; Olympus, Germany). For plaque quantification, similar and comparable histological areas were selected, focusing on adjacent positioning of the hippocampus and the whole cortical area.
Data were analyzed utilizing the comparative Cycle threshold (Ct) method (ΔΔCt), where the actin transcript level was utilized to normalize differences in sample loading and preparation. Each sample (n=4-5) was analyzed in triplicate, and the results represented the nfold difference of transcript levels among different samples.
Statistical analysis. Data are expressed as the mean ± Standard Error of the Mean (SEM) from at least 4-5 samples. Data analysis was conducted using GraphPad Prism ver. 6 statistical software. Means were compared with two-way Analysis of Variance (ANOVA) and post hoc analysis. Comparisons between groups were performed by the unpaired Student t test for independent samples. Statistical significance was considered when p values were <0.05. Statistical outliers were performed out with the Grubs test and were removed from analysis.
In addition, partial correlation controlling for group were calculated using SPSS 19.00, between the following variables: DI, Time in open arms, Locomotor activity, Amyloid plaques, SOD1, 4-HNE, DNMT1 and HDAC2. Partial correlation coefficients were calculated using the data from number of Amyloid plaques obtained with S-Thioflavin staining; behavioral parameters from NORT, EPM and OFT; protein expression for SOD1 and 4-HNE; and gene expression for Dnmt1 and Hdac2.